Responsible AI Governance and Organizational Performance in Financial Institutions: A Structured Evidence Review and Governance-to-Value Conversion Framework

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Artificial intelligence (AI) is increasingly embedded in financial institutions, yet the organizational mechanisms linking AI capability to durable performance remain insufficiently specified. This study examined 18 purposively selected authoritative regulatory, supervisory, standards, and policy documents issued from 2011 through September 2026. The coding framework covered eight mechanisms: accountable ownership, data governance, validation and testing, transparency and explainability, human oversight and competence, lifecycle monitoring, third-party governance, and consumer fairness. Documentary patterns were assessed using document-level frequencies, descriptive temporal comparison, pairwise co-occurrence analysis, and negative-case analysis. Lifecycle monitoring was present in 16 of 18 documents (88.9%). Accountable ownership and validation and testing each appeared in 15 documents (83.3%), while data governance, human oversight and competence, and third-party governance each appeared in 14 documents (77.8%). Third-party governance increased descriptively from 57.1% of documents published through 2023 to 90.9% of documents published from 2024 onward. Consumer fairness remained the least prevalent mechanism (61.1%) and was unevenly integrated across prudential, operational, and consumer-protection mandates. The Governance-to-Value Conversion framework identifies four theoretically plausible pathways through which governance may contribute to performance. Governance is most likely to be enabling when control intensity reflects use-case materiality, controls extend across the AI lifecycle, and performance evidence feeds back into managerial decisions. The article offers testable propositions, a practical scorecard, and priorities for causal research.